AN EPISTEMOLOGICAL MODEL
And an Examination of the Inference
From Direct knowledge to Knowledge of Objects
Second-Year Paper
Claremont Graduate School
Epistemology
Erick Nelson
June 1, 1983
(Text re-entered December, 2007)
INTRODUCTION
In preparation for this paper, I read variety of works outlining general
epistemological theories, and articles relating to issues of debate within them.
Because of the inter-connectedness of much of the subject, and because I noticed
complex assumptions underlying even the most innocent statements, I felt that I
must try to develop a plausible epistemological model of my own. In this way, I
might find the proper place for each of the issues under consideration, and connect
them in an appropriate way. With this accomplished, I will attempt to deal with one
of the pivotal issues within the model in a deeper way.
The limitations of this paper in size keep me from attempting to provide a more
comprehensive defense of the model which is outlined.
The basic features include:
1.
2.
3.
4.
a correspondence theory of truth
a definition of knowledge in terms of belief-for-good-reasons
a foundational approach to epistemological levels
separate inferences from direct knowledge to knowledge of objects, objects to
other minds, and other minds to public knowledge
5. acceptance of direct knowledge
6. rejection of reductionism
The specific question considered in the second part of the paper will be that of the
inference from direct knowledge to knowledge of objects. How can direct knowledge
guarantee objective knowledge without deductively entailing it? I will utilize C.I.
Lewis’ concept of “qualia”, and develop it along partially-similar lines in order to
make “direct knowledge” clear, and to argue for the evidential validity of direct
knowledge. Along the way, the defects in solipsism and phenomenalism will be
made explicit.
Three points which I believe are of particular interest are: (1) the idea of
epistemological levels, (2) treatment of the Reductionist Fallacy, and (3) the Principle
of Converging Evidence.
AN EPISTEMOLOGICAL MODEL
Knowledge
Consider the following statements, each of which I would say that I “know”:
I am now typing on a typewriter.
I got this typewriter in 1967.
I have known my wife for four years.
The sun is larger than the moon.
Light travels faster than sound.
Lincoln was shot at Ford’s theatre.
Let any assertion by “S.” If I know that S, then:
(a) I believe that S.
(b) I have good reasons for believing that S.
(c) S is true.
Without (a), I am not asserting S, and so it is not “my” knowledge. [1] I have a
“right” to know S, but have not made use of my good reasons.
Without (b), I have merely a lucky guess.
And without (c), my reasons aren’t good enough, for my belief is false. [2]
This three-fold model is a traditional one that has been put forth by a variety of
philosophers. Two criticisms have been made of it: first that one must further
believe that S on the basis of one’s reasons, and that, secondly, if one makes the
rule for good reasons strong enough to ensure that S is true, condition (c) is
redundant.
So, we will simplify our rule to state: “I know that S iff I believe that S for good
reasons which guarantee that S is true.” In order for this new rule to be intelligible,
there are four questions that must be considered:
1.
2.
3.
4.
What does it mean for S to be true?
What constitutes a good reason for S?
How can a good reason guarantee that S is true?
Which propositions can we in fact know?
We will take these up in turn.
Truth
What does it mean that S is true? Common usage would indicate that any sentence
asserting some fact in the world is true whenever that fact “is the case.” The fact
are there, waiting to be discovered. There is a simple correspondence of assertion
and state-of-affairs in the world. The problem with a correspondence theory arises
when, in dealing with epistemological issues, one tries to “match” up one’s way of
finding out about the world with the world itself. How can one get “outside” one’s
means of experiencing the world to see if it is “like” the world as it is?
This notorious problem has led to an abandonment of the correspondence theory in
favor of a coherence theory or pragmatic theory of truth. These in turn have been
roundly criticized because, at bottom, they do not preserve the intuition that we are
trying to find out about the world, not about our experiences – that is, they do not
ensure correspondence.
For example, there may be (and probably are) more than one consistent theory;
consistency is at best a “negative” proof (that is, an inconsistent theory is false). [3]
Similarly, more than one theory may “work”, and so falls under the same criticism.
It seems on the face of it that the plausibility of coherence/pragmatism rests in the
conditional: “if true, then consistent/workable.” But it is obvious that the reverse
does not hold.
We will bypass these theories, and return to a new correspondence definition of
truth. Some quick definitions will help us:









Symbol – something that “stands for” something else
Proposition – that which is expressed by a declarative sentence. [4]
Intension – syntactical term: equivalent to the proposition expressed by a
sentence. [5]
Extension – semantical term: the state-of-affairs asserted by a term or
sentence. [6]
Fact – semantical term: a true proposition. [7]
Semantical Vehicle – anything which can be made to bear semantic value.
[8]
Correspondence – the “standing for” of a symbol with its referent. (not
inherently a “copy” theory) [9]
L-true – analytically true in a language by virtue of its internal rules.
F-true – correspondence which holds between proposition and its
extension in the world.
We are committed here to a theory of truth, then, that applies symbols to the world
(extension), recognizes rules that relate symbols (intension) and makes F-truth a
relation between semantical vehicles and states-of-affairs.
Levels of Good Reasons
Let us consider that proposition “Lincoln was shot at Ford’s Theatre.” This is an
historical assertion which implicitly denies an array of inconsistent propositions (e.g.
he didn’t live forever, he didn’t always avoid the theatre, the wound in his head was
not caused by slipping on a banana-peel in Peoria, etc.) – that is, propositions
inconsistent with its truth. It is either true or false, depending entirely upon “what
happened that night in Ford’s Theatre.”
How can we ascertain the “semantical value” of this statement? It seems that it first
depends upon historical evidence, which presupposes a standard by which to
evaluate evidence, which presupposes testimony, which presupposes human minds
that lie or tell truth, that presupposes some knowledge of these minds, that
presupposes some knowledge of the world of “bodies” by which we ascertain that
other minds exist, that presupposes good grounds for believing that bodies exist.
We recognize that there are different epistemological levels, that the more primary
levels must be solid before the higher complex levels can be considered. This is the
notion of epistemological primacy, and its clear delineation can save us from endless
fallacies of circularity (for example, the philosopher who tries to show that objects
exist by an appeal to the testimony of other people). This has been called a
“foundationalist” model (as opposed to a consistency model).
If every knowledge-claim must be accompanied by a “good reason”, what must be
the reason for this “good reason”, and the reason for that, in an infinite regress?
Either this process must go on indefinitely (linearly or circularly), or it must come to
an end. [10] If it comes to an end, it must do so in what philosophers have called
“direct knowledge.”
Direct knowledge does not rest on evidence, but provides evidence for knowledge of
later (higher) epistemological levels. That is why we did not define knowledge in
terms of “evidence” (which could not apply to direct knowledge), but only “good
reasons” (which is sufficiently weak to accommodate direct knowledge). Thereby,
knowledge is either based upon evidence or is direct, and if based on evidence, there
is a relationship of evidence such that it all is ultimately grounded in direct
knowledge.
Level 1: Direct Knowledge
What is it that is known directly? Naïve realism would suggest that objects
themselves are known directly, yet this has been criticized by the “argument from
illusion” in various forms. [11] The point behind this argument is that it is possible
make all sorts of mistakes about objects, and “direct knowledge” rules out the
possibility of error.
If we are not in direct contact with the world, i.e. if we have no direct knowledge of
objects, what constitutes such knowledge, and of what do we have direct knowledge
(DK)? It is thought that we have DK of sense-data, qualia, or “appearances” which
are subjective and infallible. But there are problems with each of these
characterizations. Linguistic philosophers (e.g. Ryle) have pointed out that we have
no language to describe these “events”, and that the language which is often used in
an attempt to explicate these concepts is parasitic upon the object-language of the
physical world. For example, I don’t see a “red patch” sense-datum, I see an apple.
I don’t feel a bunch of resistant-to-pressure qualia, I feel a table. Whether they are
right in this will not necessarily bear on the issues at hand; even though language is
primarily used as a public vehicle, this does not mean there are no other uses for it.
Indeed, we “hear” a sound as well as hear a train. We do “feel” a resistant-to-touch
something as well as feel a table.
An example of a “private” experience which is accommodated by language is “I feel a
pain.” For Ryle and others [12] the argument shifts from the necessity for language
to a feature of knowledge which they claim makes knowledge of pain impossible –
and that feature is that there must be possibility for error for something to qualify as
“knowledge.” How are we to evaluate this position? First, if it is to be accepted, we
must change our definition of knowledge formed at the beginning of the paper. I
would not become “I know that S iff I believe S for good reasons which guarantee
that S may be true.” Or, “ … iff I believe S for good reasons which support S
although S may be false.” It seems as if this leaves open the door for “false
knowledge” which would be counter-intuitive. I would wish to deny this restriction of
knowledge proposed by Ryle and others, and maintain the definition as stated in the
beginning. If this is merely a quibble over words (whether we call pain “knowledge”
or not), then we may invent some words which we can substitute for the offending
word which will describe certain knowledge (which, after all, we are trying to
explicate).
I will give examples of good candidates for “direct knowledge” or “certain
knowledge”, which should be epistemologically primary. I will not argue at length for
them, but save a deeper treatment for other issues.
(1) I know that I exist. This doesn’t need to be inferred from “I think”, and
needn’t be associated with the Cartesian program. Hume clamed not to find
a referent for “I” other than the sensations that were being experienced at
the time. He found no “subject.” Others have concurred, citing the difficulty
in ascertaining just what makes up a subject. [13] I will define the “I” as the
“standpoint of consciousness”, which I intuitively and certainly recognize as
something beyond neutral sensations. (I do not depend upon memory for
identity but upon self-consciousness.)
(2) Principle of Identity: that is, “A is A and not –A.” This is not, of course,
something that is known about “the world” other than about thought itself.
It is the primitive logic that is presupposed by any symbolic function. Even
in multi-valued logics, the identity of terms and rules follows this principle; it
is unavoidable. If symbols are to be functional, there must be some form of
connectives and rules of use (e.g. “and”, “or”, etc.). What they are may be
conventional.
(3) Givens: We may take something like Lewis’ “qualia” as givens, without
which we would have no raw data for knowledge. These would include the
awareness of “sensations” of sight and touch, sound and smell and taste,
“inner” sensations like pain and pleasure, and current thoughts. Memory will
be somewhat more problematic [14] in that recent memory may function as
a given, yet memory in general functions as a means for connecting qualia,
and may be faulty.
I recognize that there are many problems with the characterization of direct
knowledge which has been laid out. All three of these categories of DK have been
questioned and criticized, and a more complete theory of knowledge which seeks to
utilize DK must defend them. However, the program of this paper is to lay out the
issues in a systematic and comprehensive way, and then to concentrate on ONE
problem which is central to this scheme.
Level 2: Objects
If we can accept the validity of DK as stated above, we must find a way of inference
to enable us to know other things by way of them. Russell put it this way: “Can the
existence of anything other than our hard data be inferred from these data?” [15]
The question can be put: I know that I exist; and that what is, is, and what isn’t,
isn’t; and that I have certain experiences which I can recognize. Are they the signs
of something else, and if so, what?
There are a variety of approaches to this problem that have been tried. Locke
separated primary qualities (properties of objects) from secondary qualities
(properties of our experience). Hume questioned the validity of any inference from
one thing to another. Noting that there is no deductive inference from our
experiences to their “sources”, induction was attempted. But that itself was
inapplicable, because it relies upon the recognition of similarities, which itself is
impossible because we cannot climb outside of our experience to see what the
similarities are. Ayer characterizes the results of this process by saying that it
bolsters the skeptical argument that there is no such inference capable of giving us
inferential knowledge. [16]
An attempt to overcome this problem gave rise to phenomenalism, which claims that
all “object” statements can be translated into “observation” statements, that is, DK
statements – and that is what the object statements refer to. Mill called objects
“permanent possibilities of sensation.” There are difficulties with this position which
will be taken up below.
If there is to be inferential knowledge at all, such that we can add to and expand our
knowledge, and such that our DK can count as signs for things not directly known
(e.g. Lincoln was shot at Ford’s theatre), and if simple induction will not give us this
inferential knowledge, what can guarantee that our “good reasons” will provide
knowledge? What kind of guarantee other than deductive entailment can there be?
This is the crucial problem (it seems to me) of knowledge.
Level 3: Other Minds
If Level 2 can be conquered, we can have hope for Level 3. The traditional
explanation of the knowledge of other minds has been that of analogy: we know
(through DK) our own feelings and thoughts (not necessarily all of the, but many of
them); we know the properties of physical objects; we recognize the behavior of
certain bodies which resemble our own and act in systematic, complicated ways.
[17]
By way of analogy, we assume that they feel and think in similar ways as we do. (It
is interesting to note that children usually make the mistake of over-applying this
rule, rather than failing to apply it, in attributing personality to animals and
inanimate objects). This argument has been criticized on two levels: (a) It is
claimed that the “inductive” basis for this generalization is based on one single case,
and is therefore deficient. (b) It is also claimed that we have no knowledge of our
own “private” states except by the same observational techniques used to determine
the state of others (e.g. Ryle, Concept of Mind).
I must leave a detailed treatment of these issues, but would answer these: (a) If
there is an inferential process which will give us real objects other than simple
induction, we may be in a good position to use it, or something like it, in reinforcing
the argument from analogy (see Plantinga, God and Other Minds, who tries to make
out a case for this argument.) (b) To say that I find out about my pain in the same
way as I find out about someone else’s pain strikes me as ludicrous. I don’t have to
look at my face in the mirror to find out if I hurt. I do have to look at your face
(among other things) to find out if you hurt. I have a private experience of my pain,
so that its character AS pain (as distinguished from gyrations) is fixed. Behaviorism
seems to commit a fallacy of conflating evidence and the thing proven-by-theevidence, which will be considered below.
Level 4: Public Knowledge
The importance of establishing that there are other minds, and that they are not
reducible to physical behavior, which is itself reducible to my sensations, has an
obvious importance in the realm of ethics. Granting reductionism, there is no reason
why I should consider anyone other than myself, or sacrifice myself for another’s
“good” unless there really is a world and people independent of myself. Where the
price of phenomenalism from the object-level to the DK-level seems small, the
combination of that reduction with the one from the minds-level to the object-level is
devastating.
Of only slightly less significance is the issue of public knowledge, cooperative venture
in arriving at truth, testimony and corroboration, and education. With the
admittance of knowledge and testimony from sources other than my own private
experience, I open myself up to inferences which will enhance my knowledge of the
world, and of myself.
It is only on this level of epistemological credentials that we may benefit from many
of the insights generated in philosophical discussion. It is at this level that evidence
becomes interesting. Legal and historiographical standards of evidence can be
established, as well as scientific modes of proof.
Objections to Model
The above model of epistemological levels can be renounced at different points. The
characterization of knowledge as belief-with-good-reason has been criticized by
Austin as making the “descriptive fallacy.” Knowing is analyzed in terms of
performance. Ayer and Danto both find arguments to show that knowledge, while
often a “performative utterance”, is not exhausted by this description: there is
descriptive meaning also. [18]
A second approach is to opt for a coherence/pragmatist model of truth, rejecting the
correspondence theory, and undercutting our notion of levels.
A third approach is to conflate the levels delineated above, and set up a new series
of levels, or abandon the distinction of levels altogether. This is done by (a) denying
that there is direct knowledge, (b) asserting that there is more direct knowledge
than our model allows, (c) collapsing the levels into each other, as in reductionism
(phenomenalism), (d) denying one or more stages of inference (from direct
knowledge to knowledge of objects, object knowledge to knowledge of other minds,
knowledge of other minds to public fact), as in Hume.
Defense of Model
A defense of our model would begin with a closer analysis of truth-theories, with a
greater explication of “a correspondence between a proposition and the world.”
Next, it would critique objections to a view of knowledge as belief-with-goodreasons. Then, it would draw out the meaning of “good reasons” by showing them
to be logical (not deductive) connections between direct knowledge and those statesof-affairs in the world of which direct knowledge is a symbol or sign. In doing this:
1. We must make sure that we have direct knowledge, and argue effectively that
we do.
2. We must show the inference from DK to objects.
3. We must show the inference (perhaps a similar one, perhaps an entirely
different one) from objects to other minds.
4. We must state the rules of evidence, given testimony’s validity, by which we
determine historical, legal, and scientific fact.
Then we may, having given an answer to “What does it mean to be true? What
constitutes a good reason for S? How can a good reason guarantee that S is true?”,
we can find an answer to “Which propositions can we in fact know?” The remainder
of this paper, now that an epistemological model has been laid out in order, will be
to deal with one of the crucial steps in defending this model. I take it that the most
crucial is the question ”Given direct knowledge, what inference must we make to
have knowledge of objects-in-the-world?” … and … “Can the combination of our
direct knowledge and this inference guarantee that a proposition about objects is
true?”
INFERENCE FROM DIRECT KNOWLEDGE TO OBJECTS
Qualia
For our model, we will use C.I. Lewis’ theory of the given in experience for our basic
description of direct knowledge. We will then try to extend it to suit our problems.
Qualia are the “given” in experience. They are found by abstracting from experience
those aspects of it that are “due to” myself:
There is, in all experience, that element which we are aware that we do not
create by thinking and cannot, in general, displace or alter. As a first
approximation, we may designate it as “the sensuous.” [19]
… certain items or aspects of the content of experience satisfy the criteria of
givenness. These are, first, its specific sensuous or feeling-character, and
second, that the mode of thought can neither create nor alter it – that it
remains unaffected by any change of mental attitude or interest. [20]
The things that describe qualia would be:









Feeling character
Unaffected by change of mental attitude
Recognizable from one experience to another [21]
Designated by “locating” it within experience (by referring to
conditions of its recurring, or other relations) [22]
Not “formless” in that it is definite [23]
Occupying no temporal spread [24]
Can be described by “looks like” [25]
Impossible to make a mistake about because it is directly intuited [26]
Taken as a sign of a property [27]
Concepts give names and interpret. Before conceptualization, qualia are without
names [28], ineffable [29] and don’t yield knowledge. [30] Verbal mistakes can be
made about qualia. [31]
A concept is the “logical intension of a term.” [32] It can be individual or common;
when individual it finally rests on “familiar imagery.” [33] It expresses the
classifying attitude, tells us that something is a “real” what, and can order any
conceivable experience. [34] We can locate two important principles which will help
us understand the relation between qualia and concepts. The first is a statement
about the given (I will call it Principle Q), the second about concepts and their
relations (called by Lewis Principle A):
[Principle Q] We should beware of conceiving the given as a smooth
undifferentiated flux; that would be wholly fictitious. Experience, when it
comes, contains within it just those disjunctions which, when they are made
explicit by our attention, mark the boundaries of events, “experiences” and
things. The manner in which a field of vision or a duration breaks into parts
reflects our interested attitudes, but attention cannot mark disjunctions in an
undifferentiated field. [35]
Principle A: It must be false, that every identifiable entity in experience is
equally associated with every other. [36]
I will elaborate on what I take to be the meaning of these two principles. In A, we
see that we are given identifiable entities to conceptualize. If they are equally
associated with each other, there will be no natural connections between them
discernible. Any connections will be arbitrary. In Q, we see that we ARE given
identifiable entities, not an undifferentiated field. On a pre-conceptual level, we can
identify, recognize, and name (in a sense), and sort qualia (according to their
recognizable properties). How this is done without concepts is not explained. We
must be able, on Lewis’ account, to use the Identity Principle (A is A and not –A)
even before actual concepts come into play. We will accept these principles.
We can sort qualia into types, which we do not confuse.





First, there are touch-qualia, which give us the “feeling” quality of hardness
(resistance to pressure), pressure, and texture (rough or smooth).
Next, there are sight-qualia, which give us the quality of brightness, patterns
(e.g. shapes), and color.
Third, we have “inner” feelings of pain, pleasure, and other sensations.
Fourth, we have some sort of recognition of our own thoughts, including
memories and imagination.
There are others: smell, taste, hearing, etc.
This thesis does not commit us to the view that we always know what qualia are
being presented, for example in very faint sounds or “funny” feelings. All that is
required is that we can identify the stronger “givens” as they are presented, and sort
them out into recognizable categories which are determined by their nature. It
should be obvious, I would think, that we never do get these categories mixed up:
that is, I know when I feel pain that it is not sight but pain; I know when I feel pain
and a rough texture how to sort them into their categories, etc. (This is not to say
that there are no borderline cases, only that there are clear ones also).
Relativity of Knowledge
Our main problem is in finding a way to move from direct knowledge to knowledge of
the objective world. That our DK is relative to our modes of experience is obvious;
this leads us to wonder whether it is “true to” the world as it is independent of our
experience, or instead “colors” it in a way that is misleading or erroneous. Lewis
puts the problem this way:
The history of philosophy since Descartes has been largely shaped by
acceptance of the alternatives; either (1) knowledge is not relative to the
mind, or (2) the content of knowledge is not the real, or (3) the real is
dependent on mind.
Kant, and phenomenalism in general, recognizes the relativity of
knowledge, the dependence of the phenomenal object on the mind, and
hence the impossibility of knowing the real as it is in itself. [37]
This is largely due to a “copy” theory, or representational theory, of knowledge:
The assumption upon which Descartes set out – the assumption of the copytheory – is that knowledge of the external world requires that the sense-quale
we apprehend should be identically present in the object perceived and in the
mind when we perceive it. … Failing this, the real object is a ding an sich, a
“something, I know not what.” [38]
Lewis’ treatment of this problem is perhaps one of his major contributions to the
theory of knowledge. He argues that the relativity of knowledge, rather than
throwing double upon the possibility of the absolute quality of the “real”, requires
“an independent character” of what is thus relative. This is a crucial point, and we
will try to trace his argument.
(1) The correspondence between our experience of the real and the real itself is not
to be thought of as a qualitatively-similar copy relationship. “Knowledge does not
copy anything presented; it proceeds from something given toward something else.”
[39]
(2) The “logic of relativity” requires that absolute knowledge be expressed in
relative terms:
If relative to R, A is X, and relative to S, A is Y, neither X nor Y is an absolute
predicate of A. But “A is X relative to R” and “A is Y relative to S”, are
absolute truths. Moreover they may be truths about the independent nature
of A. Generally speaking, if A had no independent character, it would not be
X relative to R or Y relative to S. These relative (or relational) characters, X
and Y, are partial but absolutely valid revelations of the nature of A. If we
should add “There is no truth about A which can be told without reference to
its relation to R or S, or some other such”, we should then have a very good
paradigm to the relativity of knowledge. [40]
(3) Examples of weight, size, and velocity are given to show that absolute
properties often must be described “in terms of” some relational property. The
expression of the size of Caesar’s toga in terms of the yardstick assumes that each
has a determinate size. Similarly, “the real” must, if described at all, be described in
relative terms; yet the truth of these relations must be “fixed by the objective real
character of the thing.” [41]
(4) We must view the knowledge relation as a function of two variables. There is a
systematic connection between the variables which is due to the independent object:
There is a systematic connection between masses, velocities, etc., in terms of
one relative motion and in terms of any other. So it is unnecessary for you to
try to answer in other than relative terms. The description in terms of any
relative motion, if that motion be specified, will be a sufficient description of
the nature and state of the thing. [42]
(5) We can have true, partial (not exhaustive) knowledge of the independent object
in relative terms whose nature does not imply error.
We must, further distinguish between the notion that unrecognized limitations
of the human mind would mean any deceitfulness or erroneousness of
knowledge – a failure to accord with the true nature of the real – and the
quite different notion that such limitations would mean a corresponding
degree of ignorance of reality. … our knowledge of the independent object is
veridical but partial, not that it is untrue to absolute reality. [43]
Ignorance, however great, cannot make of reality a ding an sich; it does not
vitiate such knowledge as we have, and that knowledge is of the independent
reality. [44]
Physical Objects
We will move now to a closer examination of the “relative terms” by which we
describe the independent object. This is meant to be a sketch of the ways in which
our direct knowledge might be put together to generate the physical objects in the
world. Then we will be able to look at the inferences that must be made for this
process to work.
We will limit our “hard data” to those of sight and tough, for the sake of simplicity.
Categories of touch that are groupings of “recognizable qualia” might include
hardness, pressure, and texture. Categories of sight might include brightness,
patterns (shape), and color. We must note that these two modes of experience are
independent of each other, that is, one may possess only one or the other alone, or
even alternate them (as in closing one’s eyes and touching, or standing off and
looking). It is the combination of them in a certain way which gives us the notion of
the “object.” [45]
(1) Action allows us to vary our experiences in systematic ways. We can manipulate
objects, walk around them, view them from various angles and in different lighting,
etc., to gain a more comprehensive view of their properties relative to our mode of
experience. For an entirely passive being, the only experience that would be
possible would be that which actually occurred:
For the passive being, the only possible passage of experience is the actual
one: the only continuities of reality – the only relations with any given –
would be the actual flux of experience. No object would be thicker than its
presentation. … Thus for the passive being the whole of reality would collapse
into the actual procession of the given … [46]
(2) Thought - in particular, memory – is indispensable in joining the data from the
modes of touch and sight. Thought of course initiates action. Memory must be
accountable for recognition of qualia and their relations. This of course is not to say
that memory must be infallible, but generally reliable. [47]
(3) We can get an idea of the combination of touch and sight through activity and
memory in the following way. Consider spatial relations. If we were, for instance, to
depend entirely upon sight for our estimate of size we might conclude that objects
fluctuate in size and rarely stay the same over time, because the extended region of
our visual field corresponding to any particular thing would rarely remain constant
(the thing being at various distances, etc.). However, combined with touch, we can
form a better idea of perspective and the “actual” size of objects.
With touch we can, with at least some objects, “handle” them, manipulate them,
move around them, feel their surfaces, and establish our notion of boundaries (in
terms of “yes” and “no” of touch) corresponding to certain motions of ours. In sight,
we are – broadly speaking – given the height and width of an object (that is, the
apparent height and width within our visual field), and the “real” height and width of
the object are inferred from the systematic connection between the depth (revealed
by binocular vision and the notion of perspective), apparent h and w, and apparent
size and shape determined by touch. Our memory has recorded millions of such
correlations, and can arrive at a systematic understanding of this spatial relation.
[48]
(4) Is there a clear rule which we can apply to our direct knowledge which will tell
us if we are experiencing an “object” or not? Consider something touched, but not
seen. If we can discern reflections from it, or can barely see it, we might say that it
is a “clear” object, for example, glass. But if it were absolutely unseen, and had
tactile surface and boundaries, then we would probably say that it wasn’t an object
at all, but a force field or hallucination. Perhaps we should allow for the category of
“invisible” object; it is not clear.
Consider an object that is seen but not touched. If it is touchable, but not in the
suggested sense (that is, is a picture of a cow on a billboard, but not a cow), we just
call it a picture. If it, on the other hand, is not touchable at all, if my hand passes
right through it, we would say that it is not an object, because it lacks tactile
extension; we would call it an optical illusion or hallucination, or a projection. If we
find discontinuity in our sight-touch experience (e.g. if the object passes out of
existence in terms of sight and touch), we may say either that it was not an object
or that it was an appearing and disappearing object.
Let us set up a rule which will help us to describe what it is to be an object. We will
call something an “object” which:
is pointed to by a systematic relationship between sight, touch,
memory, and action – characterized by relations of size, shape, color,
and continuity.
Such an object has been called a “logical construction” from the data of sense. The
next question is: What is the ontological status of this logical construction? May we
infer that there is a corresponding entity which possesses certain qualities or
properties expressed by these relationships found within experience, or may we not?
Phenomenalism
Phenomenalism, a form of reductionism, is a thesis which asserts that the logical
construction which is the object actually is the direct knowledge in its systematic
form. The evidence turns out to be equivalent to the thing that this known. All
language about the external world can be translated into language about direct
knowledge and relations between qualia or sense-data. In the form which accepts
valid inference (non-deductive) from qualia to other qualia, there are three major
criticisms that have been made.
(1) It is in fact impossible to really translate language about objects to an
“observation” language or protocol language, because we are inevitably dealing with
“types” of qualia, not with exact qualia as such. More telling: the conditions of
observation cannot be specified in terms of objects and relations between objects
(e.g. “not having my view obstructed”, “in proper lighting”, “within viewing
distance”, “with proper eyesight”, etc) – and it is difficult, if not impossible, to see
how this could be circumvented without begging the question (i.e. reducing the
observation to tautology of the order of “visible if visible”). ]49]
(2) Another problem with this view is the acceptance of one sort of non-deductive
inference (qualia as signs of future-experience-of-qualia) while refusing another sort
(qualia as signs of objects). C.I. Lewis seems in my opinion to go off the track at
this point. For him, [50] we have knowledge of probabilities, which is absolute
knowledge of a certain relationship holding between qualia. This assumes the
validity of memory and the ability to assign “probabilities” which seems at best
problematic. (Because of this move, he seems to fall into the phenomenalist
position.)
The inference from a memory-qualia to its “prototype” [51] of direct knowledge
seems to be analogous to the inference from qualia to the object. If this is so, he is
keeping all the difficulties of our view with the additional problems of reductionism.
(3) The most interesting problem with the phenomenalist view is that it collapses
what we know into how we know. We may call this the Reductionist Fallacy. If this
principle were to be applied in any other field of inquiry, its fallacious character
would be more easily exposed.
In a courtroom, a man could be convicted of first-degree murder, which would not
assert that he did anything, only that a gun had been found which had fingerprints
like his, that a dead body was discovered, that his next-door neighbor saw him go
into the room, etc. These things do not add up to a prison sentence, but are
evidence that (assuming a variety of things) he actually performed an action when
no one was watching. The evidence is not the action, it points to it; if sufficiently
strong, it guarantees that he must have done it.
In historiography, the statement “Lincoln was shot at Ford’s theatre” would not be
supported by evidence of written documents, etc. – it would be equivalent to those
documents. In terms of personal past, my “having known my wife for four years”
would not say anything about my past history as such, it would be equivalent to my
memories and a few (undated) pictures. The sun’s relative size to the moon would
be the few articles I might have read on the subject and vague memories from
grammar school. We could continue the list. [52]
Solipsism
The extreme form of reductionism/phenomenalism denies the existence of any
legitimate inferences from any qualia to any other that are not deductive. This
leaves us with, not only solipsism, but with a “solipsism of the moment.” Russell
comments:
… this most rigorous type of solipsist (if he exists) accepts the premise of
Descartes’ cogito, with some interpretation. What he admits can only be
correctly stated in the form: “A, B, C, … occur.” To call A, B, C, … “thoughts”
adds nothing except for those who reject solipsism. What distinguishes the
consistent solipsist is the fact that the proposition “A occurs”, if it comes in his
list, is never inferred. He rejects as invalid all inferences from one or more
propositions of the form “A occurs” to other propositions asserting the
occurrence of something, whether named or described.
… the whole of my knowledge is limited to what I am now noticing, to the
exclusion of my past and probably future, and also of all those sensations to
which, at this instant, I am not paying attention. [53]
As Russell notes, a difficulty in this rigorous solipsism is that, vitiating knowledge, it
cannot be known to be true. [54]
If Descartes’ demon were in fact a reality, such that all connections between qualia
were suspect, and such that qualia could not be taken for signs of an objective
reality; if modern skepticism’s “brain in a vat” equivalent were true – our actual
knowledge (corresponding to “the way things are”) would be tantamount to this form
of solipsism. Any epistemological model which hopes to guarantee the truth of the
external objects given direct knowledge must be able to guarantee that this
solipsism is false.
Danto, in Analytical Philosophy of Knowledge, clearly sets forth a distinction between
semantical illusion and descriptive illusion, which may be helpful. The possibility of
“semantical” illusion is the possibility that all of our semantical vehicles bear a ~
value, that there is no correspondence between them and the world. Arguments
from “illusion” trade on an equivocation between descriptive illusion (we make
mistakes in perception, etc., that can be detected) and semantical illusion (wholesale
error). Any argument which shows the possibility of error on the descriptive level
also shows the possibility for correction on that level. There are no arguments which
count directly for or against wholesale error.
Lewis sets out our choice:
If this point is established, then the only alternative to the conception that our
knowledge in general is valid, is the conception that there are no things; that
nothing exists to be known and no mind exists to know it … this alternative is
precisely the absence from experience of all order of the kind which means
significance and intelligibility. [55]
If solipsism is true (this rigorous kind), then even Lewis’ distinctions between qualia
would be fallacious. Nothing could stand for another thing (symbols would be
impossible), much less for itself. The Principle of Identity would be inarticulate. The
notion of “guarantee”, being a connection between qualia and “something” (other
qualia, objective facts, etc.) would be unintelligible.
Principle of Intelligibility
We may use signs (or symbols) to stand for anything in experience. We adopted the
Identity Principle because it is intuitively obvious. We cannot argue for it, because
its acceptance or rejection rest on its validity. In the same way, we may claim that
it is valid to use signs to signify elements of our experience which are recognizable
and abstractable. If Lewis’ conceptualization is possible, then there are connections
in our experience between qualia. I would appeal to this as intuitively known, as
direct knowledge.
Given this, we may reject solipsism as being inconsistent with our synthetic logical
principle.
Ontological Status of the Object
In Lewis, we find that the objectivity of the thing consists in there being “more” to it
than our present experience. This may be interpreted in two ways:
1. There is an object which exists independently of any observation, having a
determinate nature such that it is experience “in terms of” my modes of
experience, truly but partially (implying that, given further modes and new
experience, I might find out more about it still), OR
2. The “more” which exists is a “permanent possibility of sensation” (Mill), that
is, an anticipation of further experience. Lewis chooses the second meaning.
[56]
Norman Malcolm summarizes the argument as follows: [57]
I.
II.
III.
IV.
S has consequences.
The number of consequences of S is infinite.
The consequences of S may fail to occur.
If some of the consequences of S were to fail to occur, then there would be
a reasonable doubt that S is true.
V. If at any time there should be a reasonable doubt that S is true, then at no
previous time did anyone make absolutely certain that S is true.
So: No one did make absolutely certain that S is true.
There is a confusion in the argument at points I, III, and IV. Surely, if S entails p,
and p were to be false, then S would be false. But it is not as simple as this:



Are S’s consequences entailed by S? No. Indeed, it is difficult to spell out
the future-observations that are “consequences” of an observation.
To say that any p which is a consequence of S might fail to occur is to say
that it was not a consequence of S after all.
If p were to fail to occur, or even if it might fail to occur, how would we
tell if we were mistaken in attributing it to S or in taking S as true?
It is necessary on a “qualia-prediction” view that we only know our correlations-inmemory of past qualia (which is dubious at best), that we can assign numbers to
them, and that these numbers may fail to hold in the future, or even be applicable.
Compounded with the difficulty of translation and the seeming-impossibility of
proving just what are the consequences of an observation, this approach seems
hopeless.
The reason is that it is a victim of the Reductionist fallacy which conflates evidence
and conclusion. It is like a puppy which cannot recognize the function of “pointing”,
and instead sniffs the finer which is directing its attention to something else. The
“something more” must be more than a possibility of experience (evidence), it is that
to which all evidence is pointing. Lewis himself comes close (at least verbally close)
to this view: Intelligence is “the capacity … of transcending our individual imitations
of discrimination by indirect methods.” [58]
The idealist fails to do justice to our human power to transcend, by indirect
methods, limitations of direct experience. He should be careful not to deny,
by implication, that the blind man can believe in color. In a sense, the blind
man does not know what he believes in; nevertheless he meaningfully
believes in something that he can neither perceive nor imagine. [59]
But if the idealist puts his challenge in the form, “How can we know there is a
kind of reality we cannot know?” the different significance of the word “know”
in its two occurrences needs to be considered. Mr. Russell has pointed out
that we know there are numbers which nobody will ever count. … To know
that there are numbers not thus known, is to know a principle of the relation
of every number to others. [60]
… the independence of reality means the transcendence by reality of our
present knowledge of it; it means that I can ask significant questions about
my object which have an answer when that answer is something which I
cannot give. [61]
Are there laws which connect present and past “experience” with future
“experience”? If object-language cannot be translated into direct-knowledgelanguage verification experiences, if in fact it is not the experiences themselves but
broad categories of types of experience, and if the Reductionist fallacy is a genuine
fallacy, we see that there is something wrong with phenomenalist approaches to
knowledge.
Are there questionable probabilities holding between qualia, or are there things with
natures of their own which exemplify lawlike behavior? We have three models to
choose from:
1. Solipsism: there is no knowledge, experience is unintelligible.
2. Reductionism: translations cannot be done, laws cannot be stated, evidence
and things are collapsed.
3. Direct Knowledge yields knowledge about objects, whose independent nature
yields true, partial information to us by way of our modes of experience.
Objects can be seen and touched in certain ways; there is more to them than
meets the eye and hand.
If these are the three possibilities (I do not see more), and the first two are
impossible to hold, the third must be valid (or at the very least, infinitely better than
the other two). What is the principle that allows us to infer these objects as
independent entities?
CONVERGING EVIDENCE
This principle of Converging Evidence states:
Evidence from independent sources may compound the support for a
proposition, approaching certainty.
Example 1 – Compound Probability
This is derived from simple compound probability. Take propositions S and T; say
their probabilities are each ¾.
The probability of S and T, together, is:
P(S & T) is (PS)(PT) =
(3/4 x ¾) = 9/16 (roughly one-half).
The probability of S OR T is:
1 – [(1 – PS) x (1 – PT)] =
1 – (¼ x ¼) = 1 – 1/16 = 15/16
The probability of four propositions, each of ½, in the disjunction are also 15/16.
While evidence sources we possess do not strictly provide us with numbers, we can
see that it is a lot more difficult for many of them to be wrong than for only one.
The more independent sources, each of high probability, the better the chances of
success.
Example 2 – Triangulation
Another example is taken from “triangulation.” In three dimensions, an object’s
location may be fixed in space by three measurements. A point may be fixed in two
dimensions by two observations.
From the X position, the point may be anywhere on the X vector; from the Y
position, the point may be anywhere on the Y vector. If both measurements are
taken, there is only one point which fits both measurements. Here, the “probability”
has moved from two infinitely small quantities (it could have been anywhere on the
line) to one (certainty).
Example 3 – Legal Evidence
It can be seen in either of the above two ways: (1) One may ask, what are the
odds, given that the man’s fingerprints were on the gun, and he was seen going into
the house, and the dead body was found immediately after, and he has done this
before, and he had a motive, and … OR, (2) One might try to eliminate possibilities
by showing that evidence from independent sources allowed only one explanation for
the event. He must have done it.
Evidential Status of Direct Knowledge
We must state another principle:
Qualia constitute evidence about the objective world.
We can deduce this principle from Lewis’ own definitions: (1) We find the given by
abstracting out our own activity, (2) The Given provides knowledge with the category
“true to”, and (3) “The content of every experience is real when it is correctly
understood, and is that kind of reality which it is then interpreted to be.” [62]
Categories
We have come a long way just to show that we have knowledge of “objects.” What
shall we say about particular objects? Can we make infallible judgments about
them? Isn’t it always possible that we are mistaken? Let us take an example: a
tree. In knowing whether this is a tree, we first determine whether it is an object.
We touch it and see it. If we can’t see it because it is dark, we can say it feels like a
tree, and is probably an object (how many times have we felt a tree and found out
that it is invisible?). But, for consistency, we will wait until we can touch it and see
it. It is then an object. Does it obey laws? Certainly, it obeys the laws that classify
it as an object; this is part of its nature, expressed “in terms of” our modes of
experience. What does it take to infallibly classify it as a tree? First, what ways
could we go wrong?
(a) We might make a language mistake, not knowing English, call it a “house.”
(b) We might make a classification mistake, not knowing botany, call it a tree,
but it is really a large bush.
(c) We might think it is a tree, but it really is a fake tree, a plastic or stone tree.
(d) (fill in the blank)
Note that in all three mistakes, there was an object there to be perceived; this is not
in doubt. The only question is: how should we classify this object? If we
understand the rules for classification, and pay attention to our DK, we should be
enabled to make correct judgments. (a) occurred because of wrong word usage,
even though the “concept” expressed by the term was actually correct. (b) occurred
because we didn’t know the difference between a tree and a bush, but we could
refine our classification skills. There was nothing wrong with our data. (c) We failed
to pay close enough attention both to our data and to classification. We could have
discovered, on closer inspection that it was plastic.
Once we can have sure knowledge that there are objects, we know that they have
natures which are partially known through our direct knowledge, and that they obey
laws. More laws may be uncovered when particular objects are classified according
to more-or-less conventional rules.
SUMMARY
In this paper, I have tried to do two things: first, present a plausible model for the
theory of knowledge; and second, treat one crucial area within that model. My
model was a foundational one, having levels of epistemological primacy. The crucial
area was to define and explicate the inference from direct knowledge to knowledge
of objects.
I defined knowledge as belief for good reasons which guarantee that S is true. Two
crucial questions which arose from this definition were (a) What constitutes a good
reason, and (b) How does a good reason guarantee that S is true. In displaying the
epistemological levels, we saw that each “good reason” is some sort of inference
from one level to the next, and the sense of “guarantee”, while not deductive
entailment, must be a strong sense and might vary from level-to-level or stay the
same.
My model consisted of:
Level
Level
Level
Level
1:
2:
3:
4:
Direct Knowledge
Knowledge of Objects
Knowledge of Other Minds
Public Knowledge
I pointed out that knowledge such as “Lincoln was shot at Ford’s theatre” was of
Level 4, and thereby depended upon the possibility of knowledge at the more
primary levels.
In dealing with the inference from direct knowledge to knowledge of objects, I
assumed that we have direct knowledge, without much exposition. I found that
there were three possibilities: solipsism, phenomenalism, and independent objects
with natures (unnamed view). The first two were untenable. The third was
supported by the following:
1. Principle of Intelligibility: “We may use signs to stand for anything in
experience.” Where the Principle of Identity allowed A to equal A, this
principle allowed A to stand for x.
2. (Lewis) We abstract from our experience that which is not due to our attitude,
etc. This is called the “given”, also “qualia.”
3. These qualia exemplify systematic relations.
4. The Object is that which connects these systematic relations, to be
understood as a function of two variables f(x,y) – the subject and the object.
The object as such has a determinate nature, expressed in terms of relation.
5. Our definition for the object, then: “That which is pointed to by a systematic
relationship between sight, touch, memory, and action; characterized by
relations of size, shape, color, and continuity.”
6. Therefore, qualia can constitute “evidence” for the independently real:
“Qualia constitute evidence about the objective world.”
7. The Principle of Converging Evidence guarantees that less than certain
knowledge can converge on certainty: “Evidence from independent sources
may compound the support for a proposition, approaching certainty.”
Which propositions can we know? If the assertions listed at the beginning of the
paper, I can only that I know the first, “I am now typing on a typewriter”, because of
the rules for discerning an object, and for classifying it. The others, for the most
part, belong to higher levels, which should be considered as part of a later project.
NOTES
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
It has been questioned whether belief is always a part of knowledge. For
example, the disclaimer “I don’t know, but I believe so.” It has been pointed
out that this means – not that knowledge does not include belief – but that
the disclaimer really says, “I don’t believe-so-because-of-good-reasons, I
merely believe so.” Russell, Human Knowledge, p 154, describes the beliefcomponent of knowledge as unexceptional.
Ayer, Problem, p 35. Russell, Human Knowledge p 155. We are omitting the
senses of knowledge which are not part of “knowledge that” (e.g. “knowing
how”).
Lewis, Mind p 22. Russell, Human Knowledge p 157 and E.R. Eames,
Bertrand Russell’s Theory of Knowledge, p 153-5.
Carnap, Meaning and Necessity, p 27.
Carnap, p27.
Carnap, p 26, defines the extension of a sentence in terms of its truth-value,
yet to be consistent with his definition of extension of a term with its referent
in the world, I have altered the definition, for convenience.
There is a difference of opinion. Russell, Human Knowledge p 143 defines a
fact as something in the world (“The sun is a fact”). Danto, Analytical
Philosophy of Knowledge, p 263 says that “fact” is a semantical term, like
truth and existence, which describes the relationship between a semantical
vehicle and the world. Carnap, p 28 identifies a fact and a true proposition.
Danto, p 160. Examples of semntical vehicles and their semantic values are:
sentences (true, false), concepts (instantiated, uninstantiated), terms (refer,
fail to refer), and pictures (represent, fail to represent), p 168.
Danto, p 262: “All that is required is that we be able to associate sentences
and the world. And the terms of an association need resemble one another
no more than any signal need resemble what it signalizes. Correspondence is
a weak and an external relation, even though Correspondentialists in the past
have often supposed truth to consist in the isomorphism between sentence
and fact.”
Danto, p 27
Described by Ayer, Problem, p 87+ and many others.
Anscombe, Wittgenstein. Danto comments, p 58.
Ayer, p 48.
Russell, External World, p 72, counts recent memory as a given.
Russell, External World, p 83.
Ayer, Problems, p 76-8.
Russell, External World, p 94-6.
J.L. Austin, Philosophical Papers (Oxford; Clarendon Press), p 71. Ayer,
Problem, p 171. Danto, p 114. Danto “one-ups” Austin by calling his
criticism of our model (as the “descriptive fallacy”) the Fallacy of the Single
Function.
Lewis, p 39.
Lewis, p 66. Without qualia, there would be nothing for experiential
knowledge to be “true to.”, p 48.
Lewis, p 60, 121.
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
Lewis, p 124.
Lewis, 143.
Lewis, p 61.
Lewis, p 124.
Lewis, p 121.
Lewis, p 125.
Lewis, p 61.
Lewis, p 123-4.
Lewis, p 124.
Lewis, p 62.
Lewis, p 67.
Lewis, p 81.
Lewis, p 88, 11, 384-8 (I disagree about any experience, no matter how
chaotic, being accessible to conceptualization)
Lewis, p 59.
Lewis, p 368.
Lewis, p 154.
Lewis, p 155-6.
Lewis, p 162.
Lewis, p 168.
Lewis, p 185.
Lewis, p 187.
Lewis, p 176.
Lewis, p 180.
This is not to say that, e.g. a blind man cannot form a concept of the “object”,
just not as complete a concept. It is noteworthy that touch seems to be a
primary or ruling faculty for object-determination.
Lewis, p 141.
The problem of memory cannot be taken up here, but needs to be in a
comprehensive explication of my view.
If, for example, a person experienced one “qualia-complex” per second,
twelve hours/day, after one year he would have experienced 16 million.
Ayer, Problem, p 79, 125 (a former representative – Language, Truth, and
Logis – of a phenomenalist position)
Lewis, p 310.
Danto, his terminology.
If I have not misunderstood the phenomenalist position: If an “object” is
equivalent to my sight-DK, either my touch-DK constitutes a new object, or if
the same object, is equivalent to my sight-DK. That is, two different types of
evidence either are equivalent to different objects or to each other, making
up a complex.
Russell, Human Knowledge, p 179, 181.
Russell, p 180.
Lewis, p 378.
Lewis, p 136. “If some will contend that there can be any other kind of
“more” to the object, they must tell us how the existence of the unverifiable is
known.” (See his own refutation, quoted in footnote 60).
Lewis, p 281-2.
Lewis, p 113.
Lewis, p 180.
Lewis, p 180. (answers in footnote 56)
Lewis, p 193-4. Also, Russell, Human Knowledge, p 445+.
Lewis, p 139, 12.
BIBLIOGRAPHY
Carnap, Rudolph. Meaning and Necessity (University of Chicago) 1947.
Danto, Arthur C. Analytical Philosophy of Knowledge (Cambridge) 1968.
Ayer, A.J. The Problem of Knowledge (Penguin Books) 1956.
Eames, Elizabeth R. Bertrand Russell’s Theory of Knowledge (Braziller: New York)
1969.
Lewis, C.I. Mind and the World Order (Dover: New York) 1929.
Malcolm, Norman. Knowledge and Certainty (Cornell University) 1963.
Russell, Bertrand. Our Knowledge of the External World (Open Court: Chicago and
London) 1915.
Russell, Bertrand. Human Knowledge: Its Scope and Limits (Simon and Schuster:
New York) 1948.